VietSpeedYOLO — R420/R421 residential zone detector

YOLOv8 model for detecting Vietnam residential-zone traffic signs: R420 (Bắt đầu khu dân cư) and R421 (Hết khu dân cư). Trained on NghiMe/vietspeedyolo (Hugging Face dataset).

This release uses the YOLOv8s + offline augmentation checkpoint, which correctly distinguishes R420 (start of zone) from R421 (end of zone, red diagonal line).

Classes

ID	Label	Sign	Description
0	khu_dan_cu	R420	Start of residential area
1	ngoai_dan_cu	R421	End of residential area

Usage

Ultralytics YOLO CLI

# From local weights (after downloading from this repo)
yolo detect predict model=NghiMe/vietspeedyolo source=image.jpg save

Python

from ultralytics import YOLO

# Load from Hugging Face (requires huggingface_hub)
model = YOLO("NghiMe/vietspeedyolo")

# Or load from local path after downloading
# model = YOLO("path/to/best.pt")

results = model.predict("image.jpg", conf=0.25)
for r in results:
    boxes = r.boxes.xyxy.cpu().numpy()   # [x1, y1, x2, y2]
    classes = r.boxes.cls.cpu().numpy()  # 0 = khu_dan_cu, 1 = ngoai_dan_cu

Download and run locally

# Download best.pt from this repo's Files, then:
yolo detect predict model=best.pt source=path/to/images

Training

Base: YOLOv8s (small)
Data: 2-class YOLO dataset with offline augmentation (flip+color, rotate90+noise), see dataset card
Image size: 640

License

MIT.

NghiMe
/

vietspeedyolo